Language Modeling on WikiText-103 (Base and H + A^2 Metrics)
173.704Base ScorePythia-160M
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Pythia-160MKV bits=22026.05 | 173.704 | 19.795 | |
| Qwen2.5-0.5BKV bits=22026.05 | 94.7 | 24.498 | |
| Qwen2-0.5BKV bits=22026.05 | 84.867 | 23.108 | |
| Pythia-160MKV bits=32026.05 | 25.836 | 0.638 | |
| TinyLlama 1.1BKV bits=22026.05 | 25.248 | 14.562 | |
| GPT-2 124MKV bits=22026.05 | 24.634 | 9.072 | |
| Mistral 7BKV bits=22026.05 | 7.278 | 2.507 | |
| Qwen2-0.5BKV bits=32026.05 | 5.618 | 1.817 | |
| Pythia-160MKV bits=42026.05 | 3.73 | 0.434 | |
| GPT-2 124MKV bits=32026.05 | 3.025 | 0.709 | |
| TinyLlama 1.1BKV bits=32026.05 | 2.937 | 0.211 | |
| Qwen2-0.5BKV bits=42026.05 | 1.955 | 0.957 | |
| Qwen2.5-0.5BKV bits=42026.05 | 1.269 | 1.093 | |
| Mistral 7BKV bits=42026.05 | 0.328 | -0.281 | |
| GPT-2 124MKV bits=42026.05 | -0.134 | -0.678 | |
| Qwen2.5-0.5BKV bits=32026.05 | -0.181 | 0.91 | |
| TinyLlama 1.1BKV bits=42026.05 | -0.548 | -0.073 | |
| Mistral 7BKV bits=32026.05 | -1.522 | -0.19 |